itself is not represented in the blend; the ship image takes its place (but maintains its conceptual linkage with the state in the target input space). It is this fusion with ‘‘accommodation’’ that leads to the mental experience which Lakoff and Johnson (1980: 5) describe as ‘‘understanding and experiencing one kind of thing in terms of another.’’ A noteworthy advance of blending theory is that it allows analysts a way of describing examples in which the metaphorical image cannot be a straightforward projection of source onto target. Following Coulson (2001), Fauconnier and Turner (1998) discuss examples of the common English idiom digging [one’s] own grave, which map imagery from the domain of death and burial onto scenarios involving (nonlethal) failure of various sorts. 12 The key point from the blending analyst’s point of view is that digging a grave (one’s own or anyone else’s) does not cause death; yet instances of the idiom always refer to scenarios in which people cause themselves harm through their own actions. The causal structure of the source space is not projected into the blend; instead, the blended space contains the causal structure of the target domain, and the imagined scenario is one that would make no sense within the logic of the source input. The blending framework is obviously more powerful than CMT for describing such cases. 13 Within a blending account, Figure 8.1. Diagram of the Leonardo da Vinci’s Mona Lisa/Lewinski blend 200 joseph e. grady we can say that particular elements from each domain are counterparts (e.g., failure and death), and we can identify source and target, but we can also describe the overall structure of the resulting conceptualization, which draws in complex ways from both. The blending framework also makes it straightforward to record ways in which multiple metaphorical patterns are combined within a single complex conceptu- alization. For example, while the conventional understanding of nations as ships has nothing to say about right-left (or starboard-port) directionality, it is possible to find examples such as the following, in which the sides of the ship correspond to political orientations: ‘‘With Trent Lott as Senate Majority Leader, and Gingrich at the helm in the House, the list to the Right could destabilize the entire Ship of State’’ (see Grady, Oakley, and Couson 1999: 108). 14 This blend enlists a conven- tional association which we would ordinarily not think of as ‘‘part of’’ the Ship-of- State metaphor, and the framework allows us to treat this mapping as just another of the counterpart relations exploited by the blend. Another contribution of blending theory is that it affords an explicit means of reflecting ways in which metaphors may lead to reconceptualization of source domains (see Black’s 1955 discussion of ‘‘interaction’’ and Ricoeur’s 1978 discussion of ‘‘reverberation’’). Since blending analyses take the form of linked networks of representations, they are not directional in the same sense as CMT mappings and allow for ‘‘feedback’’ from a blended space to any of its inputs. A hypothetical ref- erence to a current military conflict as a ‘‘new Vietnam,’’ intended to frame the contemporary situation in a particular way, can also lead to new insights about the Vietnam war: see also Coulson’s (2001) blending analysis of the ‘‘Menendez Brothers virus’’ joke. A final distinction between blending theory and CMT is that the former is conceived as a description of online processing. That is, where CMT is concerned primarily with conventional patterns of association—patterns which we can think of as objects stored in long-term memory—blending is, in principle, a dynamic process. Blending scholars refer to ‘‘elaborations’’ of a blend—spontaneous de- velopments of the basic scenario constituting a blend. For instance, the Leonardo da Vinci/Lewinsky cover may suggest subsequent events such as the figure’s appearance at a congressional hearing, and so on. Since any conceptual blend depends upon identification of counterpart elements in the two input spaces, it is possible to think of conceptual metaphors (as identified within CMT) as pre- conditions for certain blends; entrenched conceptual metaphors provide one type of counterpart relationship on which blends can be based (see Grady, Oakley, and Coulson 1999). For instance, in the digging one’s own grave examples, it is clear that one original inspiration for the conceit must have been the conven- tional metaphorical association between failure and death. Following the estab- lishment of this basic connection between the two spaces, a process of selective projection leads to the more complex metaphorical structure with which we are familiar. metaphor 201 6. Computational Models of Metaphor Researchers interested in either the ongoing enterprise to model cognitive func- tions in computational terms (Artificial Intelligence) or the more specific effort to devise computational systems which accurately reproduce aspects of linguistic structure and performance (computational linguistics) have created a number of models of metaphor interpretation and production. Srini Narayanan, one of the analysts working within the Neural Theory of Language project at the University of California at Berkeley, has developed a model, for instance, which is able to gen- erate inferences about the target domains of politics and economics from linguistic input taken from newspapers and using the language of source domains such as physical motion (e.g., ‘‘the Government is currently stumbling in its efforts to im- plement the liberalization plan’’). In Narayanan’s (1999) model, source domain knowledge is represented as networks of ‘‘x-schemas,’’ representing cognitive mod- els of bodily activities, such as walking. These hypothesized schemas are under- stood as guiding bodily action, but also triggering mental simulations when the relevant concepts are evoked. The model assumes that physical domains involving such activities as walking are much more richly represented in the mind than more abstract domains such as economics. Target domains are represented as ‘‘belief nets,’’ that is, networks representing understandings about the current state of the world (or some hypothetical world). One of the central purposes of metaphor, in this view, is to take advantage of rich knowledge structures relating to physical activity in order to make it easier to think about more abstract target domains. ‘‘Since knowledge of moving around or manipulating objects is essential for sur- vival, it has to be highly compiled and readily accessible knowledge’’ (Narayanan 1999: 121). Crucially, Narayanan’s model includes ‘‘metaphor maps,’’ stored func- tions which connect source domain representations (x-schemas) to elements of target domains (belief nets). For example, one metaphor map associates a source domain event of ‘stumbling’ with a target domain representation of ‘failure’. (This is a piece of a larger map encompassing a wide range of concepts related to physical motion, representing what Lakoff 1993 has called the ‘‘Event Structure Metaphor.’’) As elaborated in the Neural Theory of Language, this framework has been extended to capture not just metaphorical associations between concepts, but also the types of relations inherent in more basic grammatical constructions, as well as the more elaborate networks of association treated by blending theory (see above). Another computational model which assumes representations of particular metaphorical mappings as part of its architecture is ATT-Meta, developed by Barnden (Barnden 2001; Lee and Barnden 2001). Like Narayanan, Barnden assumes that an important function of metaphor is to allow reasoning about such richly known conceptual domains as possession or physical motion to be applied to more abstract target domains. If the system is given a metaphorical sentence, such as I’ve 202 joseph e. grady inherited his thoughts and ideas, it first generates a set of inferences using only source domain logic (i.e., reasoning goes on within a pretence cocoon, as though the sen- tence were about literal inheritance): for instance, person A possessed something which B now possesses, A and B had a close relationship, and so on. All inferences which can apply to the target situation (in this case, one sports coach replacing an- other) are then transferred, becoming knowledge about the target domain. The projection from source to target is guided by metaphorical views implemented in the system, describing relationships between a variety of source and target domains which are conventionally linked in the minds of English speakers—in this case the idea that the logic of possession may be projected onto ideas. Other computational approaches assume no prior associations between par- ticular domains and instead seek to infer these relations based on shared properties between stored representations of large sets of concepts. Several models developed by Gentner and colleagues—notably SME (for ‘‘Structure Mapping Engine’’) and its successor MAC/FAC (for ‘‘Many Are Called, Few Are Chosen’’)—operate by seeking similarities between structural properties, either when presented with two inputs (in SME, an interpretation model) or when presented with one input and given the task of finding an appropriate analogue (in MAC/FAC, a production system) (see Falkenhainer, Forbus, and Gentner 1989; Gentner and Forbus 1991; Forbus, Gentner, and Law 1995). Veale’s Sapper system (Veale, O’Donoghue, and Keane 1995) also finds its own analogues, but looks at attributes of individual elements, as well as relational properties; the Sapper system also memorizes cor- respondences which are rich in parallels and stores them as likely candidates for future mappings (i.e., it can gradually learn a set of ‘‘conventional’’ metaphors). In effect, the models of Narayanan and Barnden are informed by theories which posit an inventory of stored conceptual mappings, while Gentner’s models focus on the capacity for interpreting and generating novel metaphorical mappings based on the perception of shared features. Veale’s model is an attempt to simulate both. While none of these early models claims to fully replicate the human capacity for creating or interpreting metaphors, such models will inevitably grow richer as the data from psychological and linguistic analyses becomes more refined, as com- putational power continues to multiply, and as findings about neurological struc- ture continue to inform the architecture of cognitive simulations. 7. Metaphor and Culture An emphasis on cognitive perspectives has led to a relative lack of attention to cultural issues within cognitive linguistic metaphor research. With certain major exceptions, researchers have been more interested in the ways that human biology and (species-wide) cognitive predispositions shape conceptualization than in the metaphor 203 ways that cultural factors shape those conceptualizations. Of course, this trend is partly in response to decades (or centuries) of special interest in metaphors as dis- tinct literary or cultural objects—characteristic products of unique societies and individual styles. Still, Cognitive Linguistics’ cognitive emphasis is complemented and enriched when scholars attend to the ways in which metaphor and culture interact. (See Dirven, Wolf, and Polzenhagen, this volume, chapter 46, for a broader discussion of the relationship between Cognitive Linguistics and culture.) One general question which presents itself to linguists interested in the rela- tionship between culture and patterns of metaphorical conceptualization is: which metaphors (if any) are culture-specific, or narrowly distributed across cultures, and which ones (if any) are universal or broadly distributed? 15 As we have seen, pri- mary metaphors are patterns that have a high likelihood of being found in any language, regardless of location, cultural affiliation, or historical period. On the other hand, there are long lists of metaphors which appear in some languages and societies but not others, and Lakoff and Johnson did not ignore this fact even in their earliest work. Here is part of their discussion of the metaphorical pattern time is money: Time in our culture is a valuable commodity Because of the way that the concept of work has developed in modern Western culture, where work is typi- cally associated with the time it takes and time is precisely quantified, it has become customary to pay people by the hour, week, or year This isn’t a necessary way for human beings to conceptualize time; it is tied to our cul- ture. There are cultures where time is none of these things. (Lakoff and Johnson 1980: 8–9; emphasis mine) Linguists may also disagree about the degree to which cultural and universal factors contribute to the genesis of a particular metaphorical pattern. There are reasons to see anger is heat, for example, as the product of the universal physiological cor- relation between the emotion and elevated skin temperature; but the humoral the- ory of emotions probably also played a role in the development, conventionali- zation, and elaboration of the pattern in Western languages. 16 There has now been a substantial amount of metaphor research in languages other than English (including work on signed languages, particularly American Sign Language; e.g., Wilcox 1993; Taub 2001). For instance, a recent special issue of Cognitive Linguistics, devoted to cross-linguistic study of terminology within the semantic domain of thought and ideas, includes a number of discussions of met- aphors in other languages (particularly, Yu’s 2003 study of Chinese). And the field is increasingly characterized by studies which treat the metaphor-culture relation- ship more centrally. Much of Michele Emanatian’s work, for example, has focused on culture-specific metaphorical patterns. She has described American models of ‘‘flexibility’’ as a highly valued trait (Emanatian 1998) and associations between food and sex that underlie linguistic usages in Chagga (a Bantu language of Tan- zania) and motivate cultural practices and taboos (Emanatian 1996). In an adjacent academic discipline, anthropologist Bradd Shore (1996) has incorporated some 204 joseph e. grady cognitive linguistic perspectives into his work on the schemas that characterize particular cultures—schemas which make up cultural knowledge and are distrib- uted in both the public world of material culture and in individual minds. A char- acteristically American schema like ‘‘modularity,’’ for example, shapes such dispa- rate institutions as the hamburger and skyscraper (each with literal part-whole structure) and the college curriculum, made up of abstract ‘‘parts.’’ 17 Cross-linguistic and cross-cultural variation, of course, can occur at any level of generality. Hiraga (1991) has observed that English and Japanese have similar lexical expressions, based on similar conceptual mappings, linking the domains of time and money; but they frame life metaphorically in terms of different sports (baseball in the case of American English, Sumo in Japanese) and have very distinct metaphors for other conceptual domains—for example, the head and mind are the seat of intentionality for English speakers, while for Japanese speakers it is the hara ‘belly’. As more comparative metaphor studies of this sort are carried out, we can expect that they will yield a clearer picture of similarities and differences between conceptual systems of people living in different cultural environments. The CMT framework has also been applied fruitfully to studies of the models operative within a single culture. Lakoff’s (1996) Moral Politics, for example, is an in-depth study of American moral and political worldviews in terms of meta- phorical models which underlie them; and Ko ¨ vacses’s (1986 and elsewhere) studies of emotion concepts discuss the role of metaphor in defining cultural models of emotion. Scholars in fields as diverse as anthropology, literary criticism, archeol- ogy, and legal studies have found the theory to be a very useful tool for analyzing the metaphorical patterns that define and permeate shared cultural understandings. 8. Additional Directions and Questions for Metaphor Research This brief essay can only provide an overview of some of the most central issues in metaphor research from the Cognitive Linguistics perspective. The following is a suggestion of several additional research areas where progress is being made, in- cluding implications of metaphor research for the understanding of other fields. 8.1. Metaphor and Attested Data Since the beginnings of the field, metaphor research within Cognitive Linguis- tics has often relied on ‘‘introspective’’ data, examples generated by the analysts themselves. While there is nothing suspect in principle about such data, which is metaphor 205 produced by native speakers and subject to confirmation by both editors and readers, the potential for refinements and additional insights to be arrived at through analysis of metaphorical language (and images, etc.) produced by ‘‘real’’ speakers and writers is self-evident. A number of research projects have worked exclusively with attested data. Cienki’s (1998) analyses of metaphorical gestures, for example, are based on examples elicited during interviews. A larger-scale project involving the participation of a number of analysts and aimed at creating a con- sistent method for the identification and analysis of metaphors encountered in literature and corpus data is directed by Steen (1999); see also Freeman’s discussion of Literary studies (this volume, chapter 45). Such research can shed light on the types of metaphors which real speakers and writers are most (or least) likely to pro- duce, the effect of context on the production of metaphors, and the discourse functions served by metaphors. 8.2. Metaphor and Neuroscience One of the most inviting frontiers for any twenty-first-century researcher inves- tigating aspects of behavior is the possibility of finding neurobiological correlates for observed patterns of mental operation. The field of cognitive neuroscience is opening new doors through which increasing numbers of metaphor analysts are bound to pass. Cognitive linguists have regularly referred to the work of neuro- scientists Antonio Damasio (e.g., 1999) and Gerald Edelman (e.g., 1991) in con- nection with hypotheses about how schemas and concepts might be represented and associated in the brain. There has also been significant research demonstrating correlations between particular types of brain activity and exposure to particular types of semantic, including metaphorical, content; see, for example, Coulson’s work on N400 effects in subjects exposed to novel metaphors (Coulson 2001; Coulson and Van Petten 2002). 18 Such work confirms that bridges, or at least the beginnings of bridges, can be built between the study of metaphorical language and the study of the brain. 8.3. Boundaries of Metaphor A number of scholars have concluded that it is difficult or impossible to make neat distinctions between metaphorical and nonmetaphorical language. For example, there is no sharp line between metaphor and cases where a category is ‘‘stretched’’ to accommodate a new item, and the difference between metaphor and literal language can be seen as a matter of degree rather than a qualitative distinction. At bottom, these difficulties arise because it is often (perhaps usually) not possible to give precise definitions for individual concepts or conceptual domains. As exam- ples of these problems, Brostro ¨ m(1994) cites uses of color terms to indicate race 206 joseph e. grady (e.g., white skin); the use of navigate in an aeronautical context; the use of disease in reference to various externally caused conditions in plants, and so on. In all these cases, she argues, there is no clear answer as to whether metaphor is involved. She also cites the following set of sentences, observing that for at least some speakers, there is no way to draw a definite line between the literal and the metaphorical among the examples: Life is a mystery / Life is a riddle / Life is a question. Instead, our judgments about the literal truth of the statements depend on how far we are willing to stretch the categories ‘Mystery’, ‘Riddle’, and ‘Question’; and such categorization gradually shades into metaphor, rather than being distinct from it. Nevertheless, there is a massive body of indisputably metaphorical examples to serve as materials for study; the ‘‘central’’ cases are clear. Continuing research will help to clarify the status of peripheral ones and the exact nature of the parameters which most accurately define the phenomenon. 8.4. Metaphor Genesis Current accounts of the origins of conceptual metaphor patterns require further confirmation and elaboration. C. Johnson’s (1999) data on developmental patterns for particular lexical items is compelling, as is Gentner’s (1988) research on chil- dren’s developing ability to understand figurative comparisons (and their prefer- ences for particular types). No evidence has yet been forthcoming, however, from these studies or others, on the exact nature of the emergence of metaphorical patterns in children’s speech or on their developing awareness of the metaphoricity of usages with which they are already familiar. It is possible, in principle, to trace the spread of a given metaphorical idiom (e.g., phrases coined by Shakespeare which have entered the general lexicon), but there is still much to be learned about the ways in which metaphorical patterns of conceptualization evolve within in- dividual minds and spread from person to person. Readers interested in pursuing the topic of metaphor further will discover inter- esting work in many areas not touched on here, including the implications of metaphor research for Western philosophy (e.g., Lakoff and Johnson 1999), for the analysis of particular conceptual domains such as mathematical thought (Lakoff and Nun ˜ ez 2000), and for readings of literature (e.g., Turner 1991; M. Freeman 1995; D. Freeman 1998); the ways in which metaphors feed into the development of grammatical systems (e.g., Brugman 1983; Svorou 1989); issues translators must face in dealing with metaphors (e.g., Mandelblit 1995); the diachronic trajectory of metaphorical patterns (e.g., Sweetser 1990); the relationship between metaphor and sound symbolism (e.g., Rhodes and Lawler 1981); and between metaphor and synesthesia (e.g., Takada, 2000). 19 The variety of questions and issues yet to be fully addressed is commensurate with metaphor’s pervasive role in thought and language. metaphor 207 NOTES I would like to thank the editors of this volume, as well as George Lakoff and Zoltan Ko ¨ vecses, for their helpful comments on an earlier draft of this article. 1. Paprotte ´ and Dirven’s (1985) collection of papers reflects the rapid adoption of Lakoff and Johnson’s approach by European linguists. 2. A number of terms have been applied to this approach—including ‘the contem- porary theory’ (Lakoff 1993), etc. For a recent and very readable book-length discussion of metaphor from a cognitive linguistic perspective, see Ko ¨ vecses (2002). For a wide- ranging set of detailed metaphor analyses in the CMT framework, see Lakoff and Johnson (1999). 3. The CMT tradition is to capitalize names of metaphorical patterns. 4. The motivation for such mappings cannot be that they frame the ‘‘unfamiliar’’ in terms of the ‘‘familiar,’’ as various accounts of metaphor have suggested, since target concepts like happiness, difficulty, and similarity are every bit as real and familiar to us as the corresponding source concepts. The advantages offered by the metaphors must be of other kinds—e.g., they allow faculties of thought and attention normally devoted to perceptual information to be applied to nonperceptual domains; see Grady (1997a) and the discussion of Narayanan’s work in section 6 below. 5. The product of such a reversal is a distinct metaphor; that is, it would be improper to say that the metaphor itself is symmetrical. Rather, in such cases, metaphorical rela- tionships between two concepts may hold in both directions, based on the same shared feature. 6. Psychologist Dedre Gentner (1988) distinguishes between metaphors based on ‘‘attributes’’ and ‘‘relations’’ (but does not consider metaphors which might be based on experiential correlation rather than any sort of shared property). 7. Grady and Johnson (2002) refer to these recurring correlations in experience as primary scenes. 8. For example, ‘‘There are no unsuccessful metaphors, just as there are no unfunny jokes’’ (Davidson 1981: 200 ). 9. See Gibbs (1994) for a helpful summary of a number of arguments and sources of evidence on this point, and Gentner (2001) for a more recent discussion of experimental data. 10. See Dean Rohrer’s cover for the New Yorker, February 8, 1999 . 11. See Turner (this volume, chapter 15) for more discussion of the basic principles and operations of blending. 12. Fauconnier and Turner refer to Coulson’s unpublished dissertation, which later became the Cambridge volume cited here. 13. CMT offers tools for analyzing some of the particular correspondences within the ‘‘digging ones own grave’’ scenario—such as the connection between death and failure, and possibly a connection between digging a hole (as though digging for a desired object) and trying to achieve a purpose. Blending theory offers a framework for combining a num- ber of distinct associations into a whole and representing inferences that do not emerge from any single metaphorical correspondence. 14. From Carol R. Campbell, ‘‘Cave Man Bill and the Doleful State of American Politics,’’ published by The Written Word, an online journal of economic, political, and social commentary. 208 joseph e. grady 15. On an even more fundamental level, of course, there are questions about the extent to which the concepts and conceptual domains linked in metaphorical patterns are themselves created and determined by culture. 16. See Ko ¨ vecses (1986) and Lakoff and Johnson (1980) for presentations of the ‘‘physiological’’ view, and Geeraerts and Grondelaers (1995) and Ko ¨ vecses (1995) for an exchange on the issue. 17. For several papers on the relationship between culture and conceptual metaphor, see Gibbs and Steen (1999). 18. A negative component in the brain’s electrical waveform, located mainly in the posterior region of the right hemisphere, peaking at about 400 milliseconds follow- ing stimulus, occurs in subjects exposed to words which require a special degree of interpretation—as in isolated words, sentence-initial words, or the punch-lines of jokes. 19. See other chapters of this Handbook for discussion of a number of these topics. REFERENCES Anderson, John M. 1971. The grammar of case: Towards a localistic theory. Cambridge: Cambridge University Press. Aquinas, St. Thomas. 1947. The Summa Theologica of Thomas Aquinas. Trans. the English Dominican Fathers. New York: Benziger Bros. Aristotle. 1996. Poetics. Trans. with introduction and notes by Malcolm Heath. Har- mondsworth, UK: Penguin Books. Barnden, John. 2001. Application of the ATT-Meta metaphor-understanding approach to various examples in the ATT-Meta project databank. Technical report CSRP-01-02, School of Computer Science, University of Birmingham, UK. Black, Max. 1955. Metaphor. In Proceedings of the Aristotelian Society, N.S. 55: 273–94. (Repr. in Mark Johnson, ed., Philosophical Perspectives on Metaphor 63– 82. Minneapolis: University of Minnesota Press, 1981) Boroditsky, Lera. 2000. Metaphoric structuring: Understanding time through spatial metaphors. Cognition 75: 1–28. Brostro ¨ m, Sofia. 1994. The role of metaphor in cognitive semantics. Lund University Cog- nitive Studies, no. 31. Lund, Sweden: Lund University Cognitive Science. Brugman, Claudia. 1983. The use of body-part terms as locatives in Chalcatongo Mixtec. In Alice Schlichter, Wallace Chafe, and Leanne Hinton, eds., Survey of California and other Indian languages 235–90. Studies in Mesoamerican Linguistics. Report no. 4. Berkeley: University of California at Berkeley. Brugman, Claudia. 1990. What is the invariance hypothesis? Cognitive Linguistics 1: 257–66. Cienki Alan. 1998. Metaphoric gestures and some of their relations to verbal metaphoric expressions. In Jean-Pierre Koenig, ed., Discourse and cognition: Bridging the gap 189–204. Stanford, CA: CSLI Publications. Clausner, Timothy, and William Croft. 1997. Productivity and schematicity in metaphors. Cognitive Science 21: 247–82. Coulson, Seana. 2001. Semantic leaps: Frame-shifting and conceptual blending in meaning construction. Cambridge: Cambridge University Press. Coulson, Seana, and Cyma Van Petten. 2002. Conceptual integration and metaphor: an event-related potential study. Memory and Cognition 30: 958–68. metaphor 209 . think of as ‘ part of ’ the Ship -of- State metaphor, and the framework allows us to treat this mapping as just another of the counterpart relations exploited by the blend. Another contribution of. instances of the idiom always refer to scenarios in which people cause themselves harm through their own actions. The causal structure of the source space is not projected into the blend; instead, the. special issue of Cognitive Linguistics, devoted to cross-linguistic study of terminology within the semantic domain of thought and ideas, includes a number of discussions of met- aphors in other languages